# R2

Published articles for R2.

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## TopoQuant: Topological Framework to Characterize Behavioral Shifts in Quantized Large Language Models

DevFeed: [TopoQuant: Topological Framework to Characterize Behavioral Shifts in Quantized Large Language Models](<https://devfeed.tech/articles/topoquant-topological-framework-to-characterize-behavioral-shifts-in-quantized-large-language-models-115071.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.00631>)

Author: Baha Rababah, Shahzeb Qamar, Lorenz Sparrenberg, Rafet Sifa, Murat Kantarcioglu, Carson K. Leung, Cuneyt Gurcan Akcora

Published: 2026-11-03T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Post-training optimization](<https://devfeed.tech/topics/post-training-optimization.md>), [quantization](<https://devfeed.tech/topics/quantization.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [LLM evaluation / benchmarking](<https://devfeed.tech/topics/llm-evaluation-benchmarking.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>)

Tags: [attention](<https://devfeed.tech/tags/attention.md>), [behavioral-shifts](<https://devfeed.tech/tags/behavioral-shifts.md>), [compression](<https://devfeed.tech/tags/compression.md>), [consistency](<https://devfeed.tech/tags/consistency.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [early-warning](<https://devfeed.tech/tags/early-warning.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [features](<https://devfeed.tech/tags/features.md>), [framework](<https://devfeed.tech/tags/framework.md>), [large-language-models](<https://devfeed.tech/tags/large-language-models.md>), [large-language-models-llms](<https://devfeed.tech/tags/large-language-models-llms.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [precision](<https://devfeed.tech/tags/precision.md>), [quantization](<https://devfeed.tech/tags/quantization.md>), [r2](<https://devfeed.tech/tags/r2.md>), [signal](<https://devfeed.tech/tags/signal.md>), [stable](<https://devfeed.tech/tags/stable.md>), [tasks](<https://devfeed.tech/tags/tasks.md>)

### AI overview

TopoQuant applies topological data analysis to self-attention weights to measure structural drift in quantized large language models. Across four model families, the authors report that drift stays stable at moderate compression, then rises sharply below 4-bit precision as attention agreement declines. Lightweight machine-learning models use topological features to predict attention consistency, with reported R² scores up to 0.85. The authors propose this as an early warning for behavioral degradation that could help assess compression limits without expensive downstream tasks; the evidence here is limited to the abstract.

### Source excerpt

arXiv:2611.00631v1 Announce Type: new Abstract: Post-training quantization (PTQ) reduces the com putational cost of large language models (LLMs), but it can introduce internal structural and behavioral changes that are not captured by traditional metrics such as accuracy and per plexity. We present TopoQuant, a topological framework that uses topological data analysis (TDA) to quantify structural drift in self-attention weights and relate it to behavioral changes in quantized LLMs. Across four LLM families and multiple quantization levels, we show that topological drift remains stable at moderate compression levels but increases sharply below 4 bit precision, closely tracking declines in attention agreement. Our results reveal a strong relationship between topological drift and attention agreement, indicating that structural changes in model parameters closely reflect behavioral degradation under quantization. We further demonstrate that topological features can accurately predict attention consistency using lightweight machine learning models, achieving R2 scores of up to 0.85. Predicting attention agreement enables efficient assessment of be havioral consistency in quantized LLMs, allowing quantization induced degradation to be detected without running expensive downstream tasks. The results indicate that topological drift pro vides an efficient early-warning signal for quantization-induced degradation and can help identify safe compression limits without extensive downstream evaluation.

## Origin Range Requests: Serving large files without range-request tuning

DevFeed: [Origin Range Requests: Serving large files without range-request tuning](<https://devfeed.tech/articles/origin-range-requests-serving-large-files-without-range-request-tuning-98271.md>)

Original publisher: [Read original article](<https://blog.cloudflare.com/origin-range-requests/>)

Author: Matthew Gumport

Published: 2026-10-23T18:22:59Z

Content type: article

Language: en

Sources: [Cloudflare Blog](<https://devfeed.tech/sources/cloudflare-blog.md>)

Topics: [proxy](<https://devfeed.tech/topics/proxy.md>), [Service fabric](<https://devfeed.tech/topics/service-fabric.md>)

Tags: [2](<https://devfeed.tech/tags/2.md>), [3](<https://devfeed.tech/tags/3.md>), [about](<https://devfeed.tech/tags/about.md>), [advertising](<https://devfeed.tech/tags/advertising.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-training](<https://devfeed.tech/tags/ai-training.md>), [bandwidth](<https://devfeed.tech/tags/bandwidth.md>), [cache](<https://devfeed.tech/tags/cache.md>), [caching](<https://devfeed.tech/tags/caching.md>), [can](<https://devfeed.tech/tags/can.md>), [files](<https://devfeed.tech/tags/files.md>), [half](<https://devfeed.tech/tags/half.md>), [large](<https://devfeed.tech/tags/large.md>), [large-files](<https://devfeed.tech/tags/large-files.md>), [performance](<https://devfeed.tech/tags/performance.md>), [pingora](<https://devfeed.tech/tags/pingora.md>), [product-news](<https://devfeed.tech/tags/product-news.md>), [r2](<https://devfeed.tech/tags/r2.md>), [range](<https://devfeed.tech/tags/range.md>), [request](<https://devfeed.tech/tags/request.md>), [small](<https://devfeed.tech/tags/small.md>), [storage](<https://devfeed.tech/tags/storage.md>), [waste](<https://devfeed.tech/tags/waste.md>)

### AI overview

Cloudflare's Origin Range Requests feature coordinates partial file delivery around fixed-size cache chunks, removing the need to tune a per-zone range size. For requested bytes already cached, Cloudflare serves them directly; it fetches missing bytes in chunk-aligned ranges, can join in-flight downloads, stores successful coverage, and returns only the client's requested range. It validates range responses, file size, and entity tags before combining bytes; when ranges cannot be honored safely, it serves the full representation. This bounds extra origin work, though requests smaller than a chunk can still fetch the whole chunk. The feature is enabled through a Cache Rule and requires an otherwise cacheable response and an origin that supports byte ranges. Cloudflare reports roughly half the p99 time-to-first-byte for R2 and a later change reduced tiered-cache traffic for R2 and similar workloads by about 6% week over week.

### Source excerpt

Naive byte-range caching can waste gigabytes of origin bandwidth on small requests. We rebuilt our cache architecture in Pingora to coordinate partial fetches automatically, cutting R2 p99 latency in half.

## Causal world models under physical parameter shifts

DevFeed: [Causal world models under physical parameter shifts](<https://devfeed.tech/articles/causal-world-models-under-physical-parameter-shifts-94873.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.29088>)

Author: Chu Lin, Yi Sun

Published: 2026-10-22T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [World models](<https://devfeed.tech/topics/world-models.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [causality](<https://devfeed.tech/topics/causality.md>), [Robotics](<https://devfeed.tech/topics/robotics.md>), [Predictive AI](<https://devfeed.tech/topics/predictive-ai.md>), [Robotics Simulation](<https://devfeed.tech/topics/robotics-simulation.md>), [Simulation](<https://devfeed.tech/topics/simulation.md>)

Tags: [benchmark](<https://devfeed.tech/tags/benchmark.md>), [control](<https://devfeed.tech/tags/control.md>), [controller](<https://devfeed.tech/tags/controller.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [dependency](<https://devfeed.tech/tags/dependency.md>), [median](<https://devfeed.tech/tags/median.md>), [mlp](<https://devfeed.tech/tags/mlp.md>), [ood](<https://devfeed.tech/tags/ood.md>), [parameter](<https://devfeed.tech/tags/parameter.md>), [pd](<https://devfeed.tech/tags/pd.md>), [r2](<https://devfeed.tech/tags/r2.md>), [recurrent](<https://devfeed.tech/tags/recurrent.md>), [robot](<https://devfeed.tech/tags/robot.md>), [synthetic](<https://devfeed.tech/tags/synthetic.md>), [world-models](<https://devfeed.tech/tags/world-models.md>)

### AI overview

In CausalWorld robot link-mass extrapolation, encoding masses explicitly in inertia and dynamics-bias terms let a known-PD world model use out-of-range masses to improve 150-step rollout accuracy: median time-averaged joint R² rose from 0.609 to 0.727, improving in all ten runs. A learned-torque variant was competitive with the strongest recurrent robotic world model in a shared 100-step forecast (median joint R² 0.755-0.768 versus 0.752), while out-of-range masses sharply degraded MLP and recurrent predictors. On a separate synthetic benchmark, a dependency mask derived from paired interventions reduced OOD RMSE from 1.769 to 0.697, near the generator-oracle mask's 0.694. The results suggest predictive effects depend on how physical parameters enter the model's dynamics mechanism.

### Source excerpt

arXiv:2610.29088v1 Announce Type: new Abstract: World models predict how embodied systems evolve under actions, but their accuracy can degrade when physical conditions move outside the training distribution. We study two components of causal modeling: which inputs a predictor may use, and how its mechanism computes their effects. In a synthetic benchmark with shortcut features, a frozen dependency mask derived from paired interventions reduces mean out-of-distribution (OOD) RMSE from 1.769 for a fully connected predictor to 0.697, close to the generator-oracle mask (0.694). For robot link-mass extrapolation in CausalWorld, with disjoint training and test mass ranges, we develop finger-local world models in which link masses enter inertia and dynamics-bias terms through explicit affine coefficients, actions act through either a learned torque network or the known PD controller, and learned residuals capture contact-related effects. Compared with a separately trained control whose mass inputs are fixed at training means, the known-PD model turns out-of-range masses into a consistent predictive gain: median time-averaged joint R2 over 150-step rollouts rises from 0.609 to 0.727, with improvement in all ten task-seed runs (exact two-sided sign test, p = 0.002). In a shared 100-step forecast, the learned-torque model is competitive with the strongest recurrent robotic world model (RWM) configuration (median joint R2 of 0.755 - 0.768 versus 0.752), whereas supplying out-of-range masses sharply degrades the MLP and recurrent predictors. These results indicate that how a physical parameter enters the dynamics mechanism determines whether its out-of-range value helps or harms prediction.

## R2-GP: Variance-Calibrated Priors for Additive Gaussian Processes

DevFeed: [R2-GP: Variance-Calibrated Priors for Additive Gaussian Processes](<https://devfeed.tech/articles/r2-gp-variance-calibrated-priors-for-additive-gaussian-processes-93628.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.26437>)

Author: Javier Enrique Aguilar, Soham Mukherjee, David Kohns

Published: 2026-10-21T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Algorithms](<https://devfeed.tech/topics/algorithms.md>)

Tags: [allows](<https://devfeed.tech/tags/allows.md>), [approximation](<https://devfeed.tech/tags/approximation.md>), [case-studies](<https://devfeed.tech/tags/case-studies.md>), [component](<https://devfeed.tech/tags/component.md>), [components](<https://devfeed.tech/tags/components.md>), [covariance](<https://devfeed.tech/tags/covariance.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [extend](<https://devfeed.tech/tags/extend.md>), [fast](<https://devfeed.tech/tags/fast.md>), [gaussian](<https://devfeed.tech/tags/gaussian.md>), [global](<https://devfeed.tech/tags/global.md>), [independent](<https://devfeed.tech/tags/independent.md>), [inference](<https://devfeed.tech/tags/inference.md>), [large](<https://devfeed.tech/tags/large.md>), [magnitude](<https://devfeed.tech/tags/magnitude.md>), [modelling](<https://devfeed.tech/tags/modelling.md>), [r2](<https://devfeed.tech/tags/r2.md>), [stat-me](<https://devfeed.tech/tags/stat-me.md>), [stat-ml](<https://devfeed.tech/tags/stat-ml.md>), [variance](<https://devfeed.tech/tags/variance.md>)

### AI overview

R2-GP calibrates prior explained variance in additive Gaussian process models using Matérn kernels. Because a component's function variance depends on both its magnitude and length scale, independent amplitude priors can overestimate total variance, especially with many components. R2-GP rescales components using their length scales and observed inputs, then assigns their rescaled variances shares of the function variance set by a global R² prior. The authors extend the approach to Hilbert space approximations for larger sample sizes. Simulations and real-world case studies report better generalisation and parameter recovery than independent variance priors.

### Source excerpt

arXiv:2610.26437v1 Announce Type: cross Abstract: We develop R2-GP, an R2D2-type prior for Gaussian process (GP) generalised additive models (GAMs) that calibrates explained variance ($R^2$) for Mat\'ern covariance kernels. The function variance for standard Mat\'ern kernels depends jointly on the effect magnitude and length-scale. As such, smooth and rough components with identical amplitudes contribute very different amounts of variation. With independent priors on component amplitudes, this variation results in variance overestimation, especially in higher dimensions. We overcome this by taking direct control over each component's expected contribution to prior explained variance. We do this by rescaling each component based on its length scale and observed inputs, and modelling each rescaled variance as a share of the function variance induced by a global $R^2$ prior. This allows R2-GP to scale well to many components. To also scale to large sample sizes, we extend R2-GP to Hilbert space approximation methods, a framework for kernels exploiting their spectral decompositions for fast and efficient inference. Our simulations and real-world case studies demonstrate superior generalisation and parameter recovery over independent variance priors.

## Stop renting access to your own files

DevFeed: [Stop renting access to your own files](<https://devfeed.tech/articles/stop-renting-access-to-your-own-files-88482.md>)

Original publisher: [Read original article](<https://blog.devops.dev/stop-renting-access-to-your-own-files-a379aa3f6123?source=rss----33f8b2d9a328---4>)

Author: Ahson Shaikh

Published: 2026-10-16T09:12:31Z

Content type: article

Language: en

Sources: [DevOps.dev](<https://devfeed.tech/sources/devops-dev.md>)

Topics: [Self Hosted File Indexing and Searching](<https://devfeed.tech/topics/self-hosted-file-indexing-and-searching.md>), [dropbox](<https://devfeed.tech/topics/dropbox.md>), [on-prem](<https://devfeed.tech/topics/on-prem.md>)

Tags: [business](<https://devfeed.tech/tags/business.md>), [drive](<https://devfeed.tech/tags/drive.md>), [dropbox](<https://devfeed.tech/tags/dropbox.md>), [files](<https://devfeed.tech/tags/files.md>), [google](<https://devfeed.tech/tags/google.md>), [google-drive](<https://devfeed.tech/tags/google-drive.md>), [google-one](<https://devfeed.tech/tags/google-one.md>), [object-storage](<https://devfeed.tech/tags/object-storage.md>), [onedrive](<https://devfeed.tech/tags/onedrive.md>), [r2](<https://devfeed.tech/tags/r2.md>), [s3](<https://devfeed.tech/tags/s3.md>), [self-hosted](<https://devfeed.tech/tags/self-hosted.md>), [self-hosting](<https://devfeed.tech/tags/self-hosting.md>), [server](<https://devfeed.tech/tags/server.md>), [storage](<https://devfeed.tech/tags/storage.md>), [yourdrive](<https://devfeed.tech/tags/yourdrive.md>)

### AI overview

yourdrive is a self-hosted web drive that puts a familiar file interface over object storage in an account the user controls. It runs in Docker on a Linux server and supports AWS S3, Azure Blob Storage, Cloudflare R2, Wasabi, MinIO, and other S3-compatible services. Browsers upload directly to the bucket, so files do not use the server's disk. The article contrasts its per-server licensing with per-user plans: Community is free for up to 3 users, Team costs $119/year for up to 25, and Business is listed at $599/year for up to 250, with SAML SSO, audit log, and groups described as coming soon. It presents this as a lower-cost option for small teams with substantial storage, while acknowledging that a household under Google One's 2 TB cap may pay less with Google. Users must maintain a Linux server; the product is a web app rather than a desktop sync client or office suite.

### Source excerpt

Why I built yourdrive, a self-hosted drive that keeps your photos, videos and documents in a bucket you already own. Count the drives you pay for. A personal Google plan for the family photos. A Dropbox seat for every person on your team. Maybe a OneDrive that came bundled with something else and now holds files you forgot about. Each one charges you every month, per person, for access to files you made. Stop paying and the files sit behind a login you no longer control. That is not storage. That is rent. I built yourdrive because I wanted the opposite: a drive that looks and feels like the one you already use, but runs on your own server and keeps every file in your own bucket. No per-seat fees. No middleman holding the keys. A drive that's actually yours. The quiet problem with rented drives Google Drive is fine for one household. Extra storage comes through Google One, at $99.99 a year for 2 TB. But that is a personal plan with a hard cap, and the moment a team needs a shared drive, the pricing switches to seats. Team drives are priced per person, not per file. Dropbox Standard costs $15 per user per month. Box Business costs $15 per user per month on an annual plan. Hire someone, and the bill goes up whether they upload one file or ten thousand. That model hides three costs: The bill tracks headcount, not storage. A 25-person team pays $4,500 a year on either plan, before anyone uploads a file. Your files live under someone else's terms. Prices, limits and features change on their schedule. You adapt. Leaving is a project. Moving years of photos and documents out is slow enough that most people never do it. Meanwhile, plenty of teams already pay for object storage: an S3 bucket for backups, an R2 bucket behind a website. It is cheap, durable and in their own account. What it lacks is a friendly drive on top, one a designer or an accountant can use without touching a terminal. What yourdrive is yourdrive is a self-hosted drive for photos, videos and documents. Yo

## Cloudflare K2 Builds Event Streams on R2 Object Storage, at a Second of Produce Latency

DevFeed: [Cloudflare K2 Builds Event Streams on R2 Object Storage, at a Second of Produce Latency](<https://devfeed.tech/articles/cloudflare-k2-builds-event-streams-on-r2-object-storage-at-a-second-of-produce-latency-83967.md>)

Original publisher: [Read original article](<https://www.infoq.com/news/2026/10/cloudflare-k2-serverless-streams/>)

Author: Steef-Jan Wiggers

Published: 2026-10-09T05:06:00Z

Content type: news

Language: en

Sources: [InfoQ](<https://devfeed.tech/sources/infoq.md>)

Topics: [Event-Streaming](<https://devfeed.tech/topics/event-streaming.md>), [Apache Kafka](<https://devfeed.tech/topics/apache-kafka.md>), [Amazon S3](<https://devfeed.tech/topics/amazon-s3.md>), [Cloudflare Workers](<https://devfeed.tech/topics/cloudflare-workers.md>)

Tags: [amazon-msk](<https://devfeed.tech/tags/amazon-msk.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [architecture-design](<https://devfeed.tech/tags/architecture-design.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [cloudflare](<https://devfeed.tech/tags/cloudflare.md>), [cloudflare-k2-serverless-streams](<https://devfeed.tech/tags/cloudflare-k2-serverless-streams.md>), [development](<https://devfeed.tech/tags/development.md>), [devops](<https://devfeed.tech/tags/devops.md>), [event-driven-architecture](<https://devfeed.tech/tags/event-driven-architecture.md>), [latency](<https://devfeed.tech/tags/latency.md>), [news](<https://devfeed.tech/tags/news.md>), [object-storage](<https://devfeed.tech/tags/object-storage.md>), [r2](<https://devfeed.tech/tags/r2.md>), [serverless](<https://devfeed.tech/tags/serverless.md>), [storage](<https://devfeed.tech/tags/storage.md>), [streaming](<https://devfeed.tech/tags/streaming.md>)

### AI overview

Cloudflare's K2 is a serverless event streaming service in public beta that builds a durable log on R2 object storage. Cloudflare reports about one second of p99 produce latency; a tester reported p99 end-to-end delivery latency of 7.5 seconds, which a team representative said was higher than expected. The article discusses K2's storage architecture, pricing, API design, and planned features.

### Source excerpt

Cloudflare has launched K2 in public beta, a serverless event streaming service implementing a durable log on R2 object storage. Cloudflare states produce latency of about one second at p99; a tester reported end-to-end p99 of 7.5 seconds, which a commenter answering for the team said was higher than expected. Consume charges match produce charges per GB. By Steef-Jan Wiggers

## AI Search is now generally available

DevFeed: [AI Search is now generally available](<https://devfeed.tech/articles/ai-search-is-now-generally-available-63113.md>)

Original publisher: [Read original article](<https://blog.cloudflare.com/ai-search-ga/>)

Author: Ashish Vinodkumar

Published: 2026-10-01T13:00:00Z

Content type: release

Language: en

Sources: [Cloudflare Blog](<https://devfeed.tech/sources/cloudflare-blog.md>)

Topics: [hybrid-search](<https://devfeed.tech/topics/hybrid-search.md>), [Embeddings](<https://devfeed.tech/topics/embeddings.md>), [Qdrant](<https://devfeed.tech/topics/qdrant.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-search](<https://devfeed.tech/tags/ai-search.md>), [birthday-week](<https://devfeed.tech/tags/birthday-week.md>), [developer-platform](<https://devfeed.tech/tags/developer-platform.md>), [developers](<https://devfeed.tech/tags/developers.md>), [pricing](<https://devfeed.tech/tags/pricing.md>), [r2](<https://devfeed.tech/tags/r2.md>), [search](<https://devfeed.tech/tags/search.md>), [vectorize](<https://devfeed.tech/tags/vectorize.md>), [visual-search](<https://devfeed.tech/tags/visual-search.md>), [what-s-new](<https://devfeed.tech/tags/what-s-new.md>), [workers-ai](<https://devfeed.tech/tags/workers-ai.md>)

### AI overview

Cloudflare's AI Search is generally available with multimodal image embeddings, OCR for scanned PDFs, and support for files up to 10 MiB. The announcement also explains its billing model, which begins November 1, 2026, and pricing by ingestion, storage, and queries.

### Source excerpt

AI Search is now generally available. It embeds image pixels directly for visual search, runs optical character recognition on scanned PDFs, accepts files up to 10 MiB, and works with any chat model. Here's what's new and how pricing works.

## Announcing Cloudflare K2: serverless event streams

DevFeed: [Announcing Cloudflare K2: serverless event streams](<https://devfeed.tech/articles/announcing-cloudflare-k2-serverless-event-streams-63115.md>)

Original publisher: [Read original article](<https://blog.cloudflare.com/cloudflare-k2-streams/>)

Author: Marc Selwan

Published: 2026-10-01T13:00:00Z

Content type: release

Language: en

Sources: [Cloudflare Blog](<https://devfeed.tech/sources/cloudflare-blog.md>)

Topics: [Event-Streaming](<https://devfeed.tech/topics/event-streaming.md>), [Apache Kafka](<https://devfeed.tech/topics/apache-kafka.md>), [Streaming](<https://devfeed.tech/topics/streaming.md>), [Cloudflare Workers](<https://devfeed.tech/topics/cloudflare-workers.md>), [Nhost](<https://devfeed.tech/topics/nhost.md>)

Tags: [birthday-week](<https://devfeed.tech/tags/birthday-week.md>), [broker](<https://devfeed.tech/tags/broker.md>), [cloudflare](<https://devfeed.tech/tags/cloudflare.md>), [data](<https://devfeed.tech/tags/data.md>), [decoupling](<https://devfeed.tech/tags/decoupling.md>), [developer-platform](<https://devfeed.tech/tags/developer-platform.md>), [event-streaming](<https://devfeed.tech/tags/event-streaming.md>), [product-news](<https://devfeed.tech/tags/product-news.md>), [r2](<https://devfeed.tech/tags/r2.md>), [serverless](<https://devfeed.tech/tags/serverless.md>), [workers](<https://devfeed.tech/tags/workers.md>)

### AI overview

Cloudflare introduces K2, a serverless event streaming service in public beta. It stores events as an ordered, durable log on R2 object storage, decoupling producers from consumers and supporting batch reads, parallel subscriptions, and long-term retention. The article explains K2's edge-oriented design, trade-offs including roughly one-second 99th-percentile produce latency, how it differs from Queues and Basin Pipelines, and ways to create and consume streams.

### Source excerpt

Cloudflare K2 is a serverless event streaming service built directly on top of R2 object storage for high-scale data movement and long-term retention. By decoupling producers and consumers at the edge, K2 enables durable, ordered log streams without the operational overhead of traditional broker clusters.

## Introducing Cloudflare Basin: an open, serverless data platform, now generally available

DevFeed: [Introducing Cloudflare Basin: an open, serverless data platform, now generally available](<https://devfeed.tech/articles/introducing-cloudflare-basin-an-open-serverless-data-platform-now-generally-available-63114.md>)

Original publisher: [Read original article](<https://blog.cloudflare.com/cloudflare-basin/>)

Author: Micah Wylde

Published: 2026-10-01T12:58:00Z

Content type: release

Language: en

Sources: [Cloudflare Blog](<https://devfeed.tech/sources/cloudflare-blog.md>)

Topics: [Apache Iceberg](<https://devfeed.tech/topics/apache-iceberg.md>), [Apache Kafka](<https://devfeed.tech/topics/apache-kafka.md>), [data lake](<https://devfeed.tech/topics/data-lake.md>), [Serverless](<https://devfeed.tech/topics/serverless.md>), [Cloudflare](<https://devfeed.tech/topics/cloudflare.md>), [Amazon S3 Tables](<https://devfeed.tech/topics/amazon-s3-tables.md>), [Developer Platform](<https://devfeed.tech/topics/developer-platform.md>), [Cloudflare Workers](<https://devfeed.tech/topics/cloudflare-workers.md>)

Tags: [2025](<https://devfeed.tech/tags/2025.md>), [2026](<https://devfeed.tech/tags/2026.md>), [apache-iceberg](<https://devfeed.tech/tags/apache-iceberg.md>), [aws-s3](<https://devfeed.tech/tags/aws-s3.md>), [birthday-week](<https://devfeed.tech/tags/birthday-week.md>), [build](<https://devfeed.tech/tags/build.md>), [building](<https://devfeed.tech/tags/building.md>), [cloudflare](<https://devfeed.tech/tags/cloudflare.md>), [data](<https://devfeed.tech/tags/data.md>), [data-platform](<https://devfeed.tech/tags/data-platform.md>), [developers](<https://devfeed.tech/tags/developers.md>), [general-availability](<https://devfeed.tech/tags/general-availability.md>), [iceberg](<https://devfeed.tech/tags/iceberg.md>), [large-datasets](<https://devfeed.tech/tags/large-datasets.md>), [object](<https://devfeed.tech/tags/object.md>), [object-storage](<https://devfeed.tech/tags/object-storage.md>), [platform](<https://devfeed.tech/tags/platform.md>), [product-news](<https://devfeed.tech/tags/product-news.md>), [query](<https://devfeed.tech/tags/query.md>), [r2](<https://devfeed.tech/tags/r2.md>), [scale](<https://devfeed.tech/tags/scale.md>), [serverless](<https://devfeed.tech/tags/serverless.md>), [sql](<https://devfeed.tech/tags/sql.md>), [storage](<https://devfeed.tech/tags/storage.md>)

### AI overview

Cloudflare announced that its serverless analytics platform, now called Cloudflare Basin, is generally available. Built on Apache Iceberg and R2 Object Storage, Basin combines event ingestion and transformation, Iceberg catalog management, and distributed SQL querying. The article describes its support for open data formats and external query engines, usage-based pricing without egress fees, and updates to scalability, table maintenance, and SQL capabilities.

### Source excerpt

Cloudflare Basin is built on Apache Iceberg and R2 Object Storage. With general availability, developers can ingest, manage, and query large datasets at scale without paying data egress fees.

## What's the best object storage for a SaaS? (September 2026 Comparison)

DevFeed: [What's the best object storage for a SaaS? (September 2026 Comparison)](<https://devfeed.tech/articles/what-s-the-best-object-storage-for-a-saas-september-2026-comparison-78703.md>)

Original publisher: [Read original article](<https://upstash.com/blog/whats-the-best-object-storage-for-a-saas-september-2026-comparison>)

Author: Josh

Published: 2026-09-23T00:00:00Z

Content type: comparison

Language: en

Sources: [Upstash](<https://devfeed.tech/sources/upstash.md>)

Topics: [Amazon S3](<https://devfeed.tech/topics/amazon-s3.md>), [Cloudflare](<https://devfeed.tech/topics/cloudflare.md>), [Comparing SaaS, PaaS, and IaaS](<https://devfeed.tech/topics/comparing-saas-paas-and-iaas.md>)

Tags: [access-control](<https://devfeed.tech/tags/access-control.md>), [aws](<https://devfeed.tech/tags/aws.md>), [aws-s3](<https://devfeed.tech/tags/aws-s3.md>), [bandwidth](<https://devfeed.tech/tags/bandwidth.md>), [blob](<https://devfeed.tech/tags/blob.md>), [cloudflare](<https://devfeed.tech/tags/cloudflare.md>), [comparison](<https://devfeed.tech/tags/comparison.md>), [cost](<https://devfeed.tech/tags/cost.md>), [distributed](<https://devfeed.tech/tags/distributed.md>), [object-storage](<https://devfeed.tech/tags/object-storage.md>), [r2](<https://devfeed.tech/tags/r2.md>), [s3](<https://devfeed.tech/tags/s3.md>), [saas](<https://devfeed.tech/tags/saas.md>), [storage](<https://devfeed.tech/tags/storage.md>), [use-cases](<https://devfeed.tech/tags/use-cases.md>)

### AI overview

This article compares Upstash Blob, Cloudflare R2, AWS S3, and Vercel Blob for SaaS file storage, weighing egress and request pricing, free tiers, CDN and regional behavior, S3 compatibility, access controls, and upload tooling. It outlines tenant file organization with bucket prefixes and signed URLs. Its examples recommend Upstash Blob for many SaaS workloads, R2 when downloads are very large, and S3 for teams already using AWS.

### Source excerpt

A comparison of Cloudflare R2, AWS S3, and Upstash Blob for SaaS use cases.

## I vibe coded my own Scrapingbee alternative and replaced my $249/mo plan with a $1/mo scraper

DevFeed: [I vibe coded my own Scrapingbee alternative and replaced my $249/mo plan with a $1/mo scraper](<https://devfeed.tech/articles/i-vibe-coded-my-own-scrapingbee-alternative-and-replaced-my-249-mo-plan-with-a-1-mo-scraper-105212.md>)

Original publisher: [Read original article](<https://levels.io/vibe-coded-own-scrapingbee-alternative>)

Author: Pieter Levels

Published: 2026-09-09T00:00:00Z

Content type: opinion

Language: en

Sources: [@levelsio (Pieter Levels)](<https://devfeed.tech/sources/levelsio-pieter-levels.md>)

Topics: [hosting](<https://devfeed.tech/topics/hosting.md>), [remote-development](<https://devfeed.tech/topics/remote-development.md>), [X (Twitter)](<https://devfeed.tech/topics/twitter.md>), [alt:V](<https://devfeed.tech/topics/altv.md>), [PaLM](<https://devfeed.tech/topics/palm.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [automation](<https://devfeed.tech/tags/automation.md>), [backblaze](<https://devfeed.tech/tags/backblaze.md>), [backblaze-b2](<https://devfeed.tech/tags/backblaze-b2.md>), [backups](<https://devfeed.tech/tags/backups.md>), [cloudflare](<https://devfeed.tech/tags/cloudflare.md>), [domains](<https://devfeed.tech/tags/domains.md>), [email](<https://devfeed.tech/tags/email.md>), [hetzner](<https://devfeed.tech/tags/hetzner.md>), [llms](<https://devfeed.tech/tags/llms.md>), [moderation](<https://devfeed.tech/tags/moderation.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [playwright](<https://devfeed.tech/tags/playwright.md>), [r2](<https://devfeed.tech/tags/r2.md>), [saas](<https://devfeed.tech/tags/saas.md>), [scraper](<https://devfeed.tech/tags/scraper.md>), [scraping](<https://devfeed.tech/tags/scraping.md>), [scrapingbee](<https://devfeed.tech/tags/scrapingbee.md>), [storage](<https://devfeed.tech/tags/storage.md>), [stuff-i-made](<https://devfeed.tech/tags/stuff-i-made.md>), [vibe-coding](<https://devfeed.tech/tags/vibe-coding.md>), [vps](<https://devfeed.tech/tags/vps.md>), [xai](<https://devfeed.tech/tags/xai.md>)

### AI overview

The author reports replacing a $249/month Scrapingbee plan with a self-built scraper costing about $1/month on a VPS. OpenSERP was blocked on the author's VPS and laptop; the author also says residential proxies from IPRoyal were frequently blocked and vanilla headless Chrome was detected. A Playwright setup using real Google Chrome in headful mode, a persistent context, slow pacing, and different residential IPs reportedly reached about 90% success, similar to or better than Scrapingbee. These are the author's reported results; the IP recommendations are not disclosed.

### Source excerpt

✨ Okay with lots of help from @javilopen and his Spanish scraping friends I've managed to vibe code my own @Scrapingbee alternative and replace my $249/mo Scra...

## All you need to build a business is free open source software a VPS an AI API and R2/S3

DevFeed: [All you need to build a business is free open source software a VPS an AI API and R2/S3](<https://devfeed.tech/articles/all-you-need-to-build-a-business-is-free-open-source-software-a-vps-an-ai-api-and-r2-s3-104822.md>)

Original publisher: [Read original article](<https://levels.io/build-business-free-open-source-vps-ai-r2-s3>)

Author: Pieter Levels

Published: 2026-07-14T00:00:00Z

Content type: opinion

Language: en

Sources: [@levelsio (Pieter Levels)](<https://devfeed.tech/sources/levelsio-pieter-levels.md>)

Topics: [Open Source](<https://devfeed.tech/topics/open-source.md>), [API](<https://devfeed.tech/topics/api.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-api](<https://devfeed.tech/tags/ai-api.md>), [api](<https://devfeed.tech/tags/api.md>), [build](<https://devfeed.tech/tags/build.md>), [business](<https://devfeed.tech/tags/business.md>), [entrepreneurship](<https://devfeed.tech/tags/entrepreneurship.md>), [free](<https://devfeed.tech/tags/free.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [open-source-software](<https://devfeed.tech/tags/open-source-software.md>), [r2](<https://devfeed.tech/tags/r2.md>), [s3](<https://devfeed.tech/tags/s3.md>), [saas](<https://devfeed.tech/tags/saas.md>), [software](<https://devfeed.tech/tags/software.md>), [tech](<https://devfeed.tech/tags/tech.md>), [vps](<https://devfeed.tech/tags/vps.md>)

### AI overview

The author argues that a business can be built using free open-source software, a VPS server, an AI API for required AI tasks, and R2/S3 file hosting. The excerpt offers this as a minimal-cost stack, but provides no implementation details or evidence for the claim.

### Source excerpt

All you really need today to build a business is 100% free open source software that charges you $0/mo a VPS server an API to do some required AI stuff some R2...

## Reversing Pickles with r2pickledec

DevFeed: [Reversing Pickles with r2pickledec](<https://devfeed.tech/articles/reversing-pickles-with-r2pickledec-92033.md>)

Original publisher: [Read original article](<https://blog.doyensec.com/2023/06/01/r2pickledec.html>)

Published: 2023-05-31T22:00:00Z

Content type: tutorial

Language: en

Sources: [Doyensec](<https://devfeed.tech/sources/doyensec.md>)

Topics: [code analysis](<https://devfeed.tech/topics/code-analysis.md>), [Data serialization](<https://devfeed.tech/topics/data-serialization.md>), [Tool](<https://devfeed.tech/topics/tool.md>), [Python](<https://devfeed.tech/topics/python.md>), [Assembly](<https://devfeed.tech/topics/assembly.md>)

Tags: [auditing](<https://devfeed.tech/tags/auditing.md>), [automation](<https://devfeed.tech/tags/automation.md>), [decompiler](<https://devfeed.tech/tags/decompiler.md>), [doyensec](<https://devfeed.tech/tags/doyensec.md>), [doyensecurity](<https://devfeed.tech/tags/doyensecurity.md>), [exploit](<https://devfeed.tech/tags/exploit.md>), [hacking](<https://devfeed.tech/tags/hacking.md>), [infosec](<https://devfeed.tech/tags/infosec.md>), [obfuscation](<https://devfeed.tech/tags/obfuscation.md>), [offensive](<https://devfeed.tech/tags/offensive.md>), [pentest](<https://devfeed.tech/tags/pentest.md>), [pentester](<https://devfeed.tech/tags/pentester.md>), [protocol](<https://devfeed.tech/tags/protocol.md>), [python](<https://devfeed.tech/tags/python.md>), [r2](<https://devfeed.tech/tags/r2.md>), [remote-code-execution](<https://devfeed.tech/tags/remote-code-execution.md>), [reverse-engineering](<https://devfeed.tech/tags/reverse-engineering.md>), [reversing](<https://devfeed.tech/tags/reversing.md>), [security](<https://devfeed.tech/tags/security.md>), [serialization](<https://devfeed.tech/tags/serialization.md>), [stack](<https://devfeed.tech/tags/stack.md>), [stream](<https://devfeed.tech/tags/stream.md>), [tool](<https://devfeed.tech/tags/tool.md>), [vulnerability](<https://devfeed.tech/tags/vulnerability.md>)

### AI overview

r2pickledec decompiles Python pickles through Radare2, supporting instructions through protocol 5. The article explains that pickle instructions mostly manipulate a stack, with a memo for retrieving earlier values, and warns that unpickling untrusted data can lead to remote code execution. In its example, the decompiler reveals a serialized API object's requests session and an auth tuple containing credentials, then locates and patches the object's base URL. Radare2 flags help identify and rename variables; the plugin's JSON AST includes offsets that can support scripted searches and edits with r2pipe. The author says packaged Radare2 versions may be old but should work for this use, and suggests installing from source if bugs occur.

### Source excerpt

R2pickledec is the first pickle decompiler to support all instructions up to protocol 5 (the current). In this post we will go over what Python pickles are, how they work and how to reverse them with Radare2 and r2pickledec. An upcoming blog post will go even deeper into pickles and share some advanced obfuscation techniques. What are pickles? Pickles are the built-in serialization algorithm in Python. They can turn any Python object into a byte stream so it may be stored on disk or sent over a network. Pickles are notoriously dangerous. You should never unpickle data from an untrusted source. Doing so will likely result in remote code execution. Please refer to the documentation for more details. Pickle Basics Pickles are implemented as a very simple assembly language. There are only 68 instructions and they mostly operate on a stack. The instruction names are pretty easy to understand. For example, the instruction empty_dict will push an empty dictionary onto the stack. The stack only allows access to the top item, or items in some cases. If you want to grab something else, you must use the memo. The memo is implemented as a dictionary with positive integer indexes. You will often see memoize instructions. Naively, the memoize instruction will copy the item at the top of the stack into the next index in the memo. Then, if that item is needed later, a binget n can be used to get the object at index n. To learn more about pickles, I recommend playing with some pickles. Enable descriptions in Radare2 with e asm.describe = true to get short descriptions of each instruction. Decompile simple pickles that you build yourself, and see if you can understand the instructions. Installing Radare2 and r2pickledec For reversing pickles, our tool of choice is Radare2 (r2 for short). Package managers tend to ship really old r2 versions. In this case it's probably fine, I added the pickle arch to r2 a long time ago. But if you run into any bugs I suggest installing from source. In